Thin films of polycrystalline cerium substituted yttrium iron garnet (CeYIG) were grown on an yttrium iron garnet (YIG) seed layer on Si and Si-on-insulator substrates by pulsed laser deposition, and their optical and magneto-optical properties in the near-IR region were measured. A YIG seed layer of ~30 nm thick processed by rapid thermal anneal at 800°C provided a virtual substrate to promote crystallization of the CeYIG. The effect of the thermal budget of the YIG/CeYIG growth process on the film structure, magnetic and magnetooptical properties was determined.
©2012 Optical Society of America
Non-reciprocal devices such as isolators [1–4] or circulators [5, 6] are important components of photonic integrated circuits . Non-reciprocal propagation of light can be achieved by incorporating magneto-optical (MO) materials into these devices. Discrete isolators and circulators have been made successfully for years using bulk magnetooptical garnets, but monolithic integration of these materials onto non-garnet substrates, e.g. Si or Si-on-insulator (SOI), has been challenging because of thermal and lattice mismatch and the formation of undesirable non-garnet phases.
The prototype ferrimagnetic garnet, yttrium iron garnet (Y3Fe5O12 or YIG) is highly transparent at near-IR communication wavelengths but has a relatively weak magnetooptical activity, so for magnetooptical devices the Y is partly substituted by elements such as Bi or rare earths which can increase the Faraday rotation by an order of magnitude. These garnets can be grown epitaxially on garnet-structured substrates and have been patterned by etching. Cerium substituted yttrium iron garnet (CexY3-xFe5O12 or CeYIG) has been widely studied [8, 9]. Bulk single crystal CeYIG can be prepared by the traveling solvent floating zone method, but the cerium substitution x was limited to 0.24 . Because of this, thin film deposition techniques have been used widely to form CeYIG with higher Ce content. Shintaku et al. [10–12] coupled light into single crystal films of CeYIG on garnet substrates and measured the extinction coefficient and Faraday rotation angle at 1550 nm. Gomi et al. [13–15] sputtered CeYIG on garnet and thermally-oxidized Si substrates. CeYIG with a Ce substitution of x = 1 (i.e. Ce1Y2Fe5O12) could be grown on garnet substrates, but on non-garnet substrates the films were polycrystalline and x was limited to 0.7 before other phases appeared [14, 16].
Recently, we demonstrated a two-step-deposition method using pulsed laser deposition (PLD) in which a thin YIG buffer layer was used as a seed layer for growth of polycrystalline CeYIG on non-garnet substrates [1, 17] and used this process to make an integrated isolator based on a ring resonator. The YIG seed layer provides a virtual substrate on which CeYIG can grow, lessening the tendency for unwanted phases such as CeO2 to form, even for high Ce contents such as x = 1. This process produced films with Faraday rotation angle of –800 deg/cm and figure of merit (FOM) of 20 deg/dB, but required processing at high temperatures during the three steps of YIG growth (550þC), a subsequent rapid thermal anneal (850þC), and then the CeYIG growth (650þC). From the point of view of integration with other on-chip optical components, it is important to minimize the thermal budget of the film deposition and annealing processes. In this work, we report the structural, magnetic and magnetooptical performance of YIG and CeYIG grown under conditions that reduce the thermal budget of each of the three process steps, to produce YIG/CeYIG bilayers on Si and SOI. The Faraday rotation angle, refractive index, extinction coefficient and FOM of polycrystalline YIG and CeYIG are reported and compared with data from other single crystal films.
2. Experimental methods
The films were deposited using PLD with a KrF excimer laser (Coherent, COMPex Pro 205) at wavelength λ = 248 nm. The targets of yttrium iron garnet (Y3Fe5O12, YIG) and cerium substituted yttrium iron garnet (Ce1Y2Fe5O12, CeYIG) were prepared by a conventional mixed oxide sintering method [18, 19] from powders (Alfa Aesar) of Y2O3 and Fe2O3 for YIG, with CeO2 added for the CeYIG target. These compounds were calcined at 1150°C for 5 hours, pressed into a 1 inch diameter, 0.5 inch thick cylinder shape, then sintered at 1400°C for 10 hours in air, which resulted in pure garnet phase targets identified by xray diffraction (XRD). During film deposition the frequency of the pulsed laser was 10 Hz, pulse duration time was 25 ns, spot size on the target was 0.5 × 2.0 mm, and energy density at the target surface was ~1.3 J/cm2 for one pulse. The distance between the target and the substrate was 60 mm. During deposition the target was rotated to ensure laser erosion over the majority of the target surface, and the substrates were rotated to make the film thickness uniform. The targets were polished by sandpaper before each deposition because the laser erosion caused changes in the surface target composition. The window of the chamber through which the laser beam entered was also cleaned before each deposition in order to maintain constant pulse energy at the surface of the target.
The YIG films were deposited on SOI substrates comprising Si (0.4 mm)/SiO2 (3 μm)/Si (220 nm) and on Si substrates (0.381 mm), where the parentheses show the thicknesses, at 100°C in 1.5 × 10−6 Torr without any added gas flow. This condition was chosen to minimize the thermal budget of the YIG deposition step because as-deposited films were amorphous, whether the substrate temperature was at 550°C or 100°C, with or without oxygen. The deposition rate was ~1.2 nm/min to form a 349 nm thick YIG film on the Si and SOI substrates.
349 nm thick YIG films were processed by rapid thermal annealing (RTA, Modular Process Tech, RTP-600S)  at various temperatures for 3 minutes in 5 standard liters per minute (slpm) oxygen flow. The substrate temperature was increased from room temperature (RT, 25þC) to the set point of up to 900°C within 30 seconds, held for 3 minutes, and decreased to RT within 4 minutes.
CeYIG was deposited onto YIG seed layers of various thicknesses which had been deposited at 100þC then annealed by RTA at 800 °C for 3 minutes. The CeYIG was grown on the seed layer at the same condition described in Ref . (i.e. at 650°C in 20 mTorr oxygen). The deposition rate of CeYIG was ~3.0 nm/min to form a ~150 nm thick CeYIG film. The set of samples can be described as Si substrate/YIG (6–136 nm)/CeYIG (~150 nm). To derive refractive index and extinction coefficient, a thicker CeYIG sample, Si/YIG (31 nm)/CeYIG (474 nm) was also formed.
The crystalline structures of deposited and annealed films were analyzed by XRD (PANalytical X’Pert PRO MPD with CuKα radiation source at the wavelength of 0.1541 nm) in ω-2θ mode. To eliminate the substrate diffraction peaks, the substrates were tilted by 2 degree during measurement. Each XRD result was offset along the y-axis in the figures for clarity. Magnetic properties were measured by vibrating sample magnetometry (VSM). Only the results of films on Si substrates were shown because there were not large differences from those on SOI substrates. The composition of a thicker deposited film measured by energy dispersive xray spectrometer (EDS, FEI/Philips XL30 FEG ESEM) was Ce0.96 ± 0.03Y1.99 ± 0.03Fe5O12.05 ± 0.12, i.e. comparable to the atomic ratio in the target. In addition to SOI substrates, the bilayers were grown on double side polished Si for Faraday rotation measurements. Transmissivity of the films grown on double side polished Si substrates was measured by spectrophotometer (VARIAN, Cary 500i) in the wavelength range of 800–1750 nm with unpolarized light perpendicular to the substrate. The surface roughness of a film consisting of SOI substrate/YIG (31 nm)/CeYIG (147 nm) was characterized with atomic force microscopy (AFM, Digital instruments Nanoscope). The Faraday rotation versus applied magnetic field of the samples were measured at a wavelength of 1550 nm with the oscillating polarizer method described in Ref , with the field and the laser beam both perpendicular to the film plane. All characterizations were carried out at ambient temperature (25þC).
3. YIG structure and properties
Figure 1 shows the diffraction data for a set of YIG samples with thickness of 349 nm, for different temperature RTA steps. Phase-pure polycrystalline YIG was obtained when the films were annealed at more than 800°C for 3 minutes. The film annealed at 750°C showed peaks from hematite (α-Fe2O3) which can contribute large optical losses. There were no crystalline peaks from samples annealed below 700°C. These patterns suggest that the most appropriate annealing condition to crystallize the YIG of this thickness was 800°C for 3 minutes.
The different RTA temperatures also affected magnetic properties as measured by VSM, Fig. 2 . The 349 nm thick YIG films showed in-plane magnetization with a saturation magnetization Ms of 154 emu/cm3 at RT, close to the value for single crystal YIG , and a hard-axis saturation field of 1.5–2.0 kOe. Comparing with the calculated shape anisotropy field of 4πMs = 1.93 kOe suggests that shape anisotropy was the dominant source of anisotropy in these films. The coercivity of ~15 Oe is small and comparable to that of single crystal films .
Figure 3(a) shows XRD data for YIG of various thicknesses after annealing at 800°C for 3 minutes. For YIG of 31 nm and below there is a broad peak indicating an amorphous or nanocrystalline phase, but above 31 nm, garnet peaks are evident.
Figure 4 gives the magnetic hysteresis loops of YIG of different thickness after the 800þC RTA. Films of 61 nm and above gave loops similar to the 349 nm thick film of Fig. 2, with 150 emu/cm3 saturation magnetization, similar to bulk values. However, as the film thickness decreased so did the saturation magnetization.
In the final YIG/CeYIG bilayer, the role of the YIG is to provide a crystallographic seed for the CeYIG growth, so its structure is more critical than its magnetic properties. An annealing condition at 800°C for 3 minutes was used for the growth of the YIG seed layer in the bilayer films, since this temperature produced only garnet peaks in the thicker YIG films.
4. YIG/CeYIG bilayer films
XRD data for bilayers of Si substrate/YIG (6–136 nm)/CeYIG (~150 nm) are given in Fig. 3(b). Garnet peaks were apparent in CeYIG when the YIG thickness was 10 nm and higher. When the YIG thickness was 6 nm, peaks from yttrium oxide were visible, though the intensity was low. These results suggest that a YIG seed layer thickness of at least 10 nm was needed to promote the growth of garnet phase CeYIG on a Si substrate. Similar behavior was observed on the SOI substrates. The out of plane lattice constant of the CeYIG grown on the 31 nm thick YIG was 12.53 ± 0.04 Å, which is similar to the lattice parameter of the 136 nm thick YIG (12.51 ± 0.03 Å).
Figure 5 gives magnetic hysteresis loops of the YIG/CeYIG, with the substrate and YIG contributions subtracted. This was done by subtracting the total magnetic moment of the Si/YIG from the net Si/YIG/CeYIG moment, then dividing the resulting magnetization by the volume of CeYIG to obtain the magnetization of CeYIG. The CeYIG saturation magnetization for YIG thickness of 31 nm reached 110 emu/cm3, slightly below the value of 120 emu/cm3 obtained for other CeYIG films of similar composition [1, 10]. These results showed that the minimum YIG thickness working as a good seed layer was 31 nm. The performance of YIG as a seed layer was more strongly correlated to its crystal structure, i.e. to the purity of the garnet phase, than to the magnitude of its saturation magnetization. Even though the 31 nm thick YIG did not have bulk magnetic properties, it still functioned as well as a 61 nm YIG seed in obtaining CeYIG with good quality.
Transmission spectra are given in Fig. 6 . The YIG films had a lower refractive index than the Si substrate and were therefore less reflective, so the Si/YIG samples had a higher transmissivity compared to uncoated Si substrates (Fig. 6(a)). However, the CeYIG had a higher optical loss and lowered the transmissivity (Fig. 6(b)).
From the transmission spectra, the refractive index and extinction coefficient of the Si substrate, YIG layer, and CeYIG layer were derived using the spectrum simulation software (SCOUT) based on conventional Fresnel equations taking account of surface roughness (Fig. 7 ). A 474 nm thick CeYIG film, Fig. 6(c), showed oscillations in the transmissivity enabling an accurate determination of the optical parameters. The same optical parameters provided an excellent fit to the observed spectra of a 369 thick CeYIG film, also shown in Fig. 7(c). These indices at 1550 nm wavelength are comparable to values reported for bulk Si and for single crystal YIG and CeYIG grown on garnet substrates (Table 1 ). The indices showed little variation with the thickness of the films and fitted well to the data of Fig. 6(a) and (b). The extinction coefficient of YIG was relatively small, so a YIG seed layer does not contribute significantly to optical losses in a device.
Figure 8 shows a three dimensional atomic force microscope (AFM) image of YIG (31 nm)/CeYIG (147 nm) on SOI substrate. The surface has 1.2 nm rms roughness and a grain size of ~10 nm. This smooth surface is important in optical devices to reduce scattering.
The saturation Faraday rotation of the YIG layer on Si was ~250 deg/cm at 1550 nm wavelength, Fig. 9(a) . This magnitude is similar to single crystal values (273 deg/cm at λ = 1100 nm in Ref , 250 deg/cm at λ = 1150 nm in Ref .). The corresponding loops for the YIG/CeYIG are given in the same figure, after subtracting the contribution of the substrates. As expected, the CeYIG has opposite sign of Faraday rotation compared to YIG. The Faraday loops have the same hard-axis shape as the out of plane magnetization loops. The bilayer has a maximum Faraday rotation of –1100 deg/cm in the case of a YIG thickness of 31 nm.
From the extinction coefficient and Faraday rotation angle, one can obtain a magnetooptical FOM which parameterizes the capability of MO material for non-reciprocal devices in optical circuits. Here, the FOM was defined as Faraday rotation angle divided by extinction (or absorption) coefficient. Table 2 shows the FOM of two-step-deposited CeYIG on Si substrate/YIG seed layer compared to a single crystalline CeYIG film on a garnet substrate, as well as to our earlier data for CeYIG on YIG deposited at higher temperatures . Other work has reported higher Faraday rotation angle in CeYIG films, e.g. –4500 dB/cm in Ref , but did not report absorption coefficient, so the FOM could not be evaluated.
One can see that two-step-deposited polycrystalline CeYIG demonstrates an order smaller FOM than single crystalline CeYIG films on garnet substrates. One possible contributing factor is the effect of Ce4+ as a result of off-stoichiometry in the film, which raises absorption and decreases the Faraday rotation angle . Grain boundaries and surface roughness may also contribute to loss. Despite this, the polycrystalline YIG/CeYIG is still an excellent material for integrated magnetooptical devices considering its ability to be grown on non-garnet substrates including SOI and Si reported here and also on silicon nitride . Comparing the FOM with that of polycrystalline CeYIG on YIG deposited at a higher temperature , the films reported here have superior FOM, even though they were deposited with a lower total thermal budget.
This work illustrates how optimized YIG films can be used as a seed layer (or virtual substrate) for the growth of CeYIG on Si or SOI substrates, with a reduced thermal budget compared to prior work. Although the YIG films deposited at 100°C were amorphous, RTA at 800°C for 3 minutes in oxygen crystallized the films into pure garnet phases. The magnetic properties, refractive index, extinction coefficient and Faraday rotation angle of the YIG films are comparable to single crystal YIG.
CeYIG grew at 650°C as a polycrystalline film on a YIG seed layer of thickness 31 nm and above. CeYIG grown on a YIG seed layer exhibited a higher Faraday rotation angle than YIG and with opposite sign. The optical absorption was larger and the saturation magnetization and Faraday rotation angle were smaller than those of single crystal films of CeYIG grown on garnet. The figure of merit was 38 deg/dB, about 9 times smaller than that reported for a single crystal CeYIG film on a garnet substrate, but higher than that we reported earlier for polycrystalline CeYIG grown with a higher thermal budget.
These results enable formation of CeYIG with useful magnetooptical properties on non-garnet substrates. This two-step deposition process can provide a route to monolithic integration of MO devices while reducing the thermal budget of the process.
TG acknowledges the Japan Society for the Promotion of Science (JSPS) Postdoctoral Fellowships for Research Abroad. CR acknowledges support of the NSF. This work made use of the shared experimental facilities of the Center for Materials Science and Engineering (CMSE), an NSF MRSEC. We thank Mr. Yu Eto, Dr. Mitsuteru Inoue, Dr. Lei Bi, Dr. Dong Hun Kim, and Dr. Gerald F. Dionne for experimental support and discussions.
References and links
1. L. Bi, J. Hu, P. Jiang, D. H. Kim, G. F. Dionne, L. C. Kimerling, and C. A. Ross, “On-chip optical isolation in monolithically integrated non-reciprocal optical resonators,” Nat. Photonics 5(12), 758–762 (2011). [CrossRef]
2. M. Levy, I. Ilic, R. Scarmozzino, R. M. Osgood, R. Wolfe, C. J. Gutierrez, and G. A. Prinz “Thin-film-magnet magnetooptic waveguide isolator,” IEEE Photon. Technol. Lett. 5(2), 198–200 (1993). [CrossRef]
3. S. Ghosh, S. Keyvavinia, W. Van Roy, T. Mizumoto, G. Roelkens, and R. Baets, “Ce:YIG/Silicon-on-Insulator waveguide optical isolator realized by adhesive bonding,” Opt. Express 20(2), 1839–1848 (2012). [CrossRef] [PubMed]
4. M.-C. Tien, T. Mizumoto, P. Pintus, H. Kromer, and J. E. Bowers, “Silicon ring isolators with bonded nonreciprocal magneto-optic garnets,” Opt. Express 19(12), 11740–11745 (2011). [CrossRef] [PubMed]
6. S. Baek, M. Dokukin, K. Yayoi, A. V. Baryshev, and M. Inoue, “Diffraction-modified Kerr rotation from patterned garnet films,” J. Appl. Phys. 107(9), 09 A923 (2010). [CrossRef]
7. C. Xiong, W. H. P. Pernice, M. Li, and H. X. Tang, “High performance nanophotonic circuits based on partially buried horizontal slot waveguides,” Opt. Express 18(20), 20690–20698 (2010). [CrossRef] [PubMed]
8. B. Stadler, K. Vaccaro, P. Yip, J. Lorenzo, Yi-Qun Li, and M. Cherif, “Integration of magneto-optical garnet films by metal-organic chemical vapor deposition,” IEEE Trans. Magn. 38(3), 1564–1567 (2002). [CrossRef]
9. S. Higuchi, Y. Furukawa, S. Takekawa, O. Kamada, K. Kitamura, and K. Uyeda, “Magnetooptical properties of cerium-substituted yttrium iron garnet single crystals for magnetic-field sensor,” Sens. Actuators A Phys. 105(3), 293–296 (2003). [CrossRef]
10. S. Mino, A. Tate, T. Uno, T. Shintaku, and A. Shibukawa, “Structure and lattice deformation of ce-substituted yttrium iron garnet film prepared by RF sputtering,” Jpn. J. Appl. Phys. 32(Part 1, No. 7), 3154–3159 (1993). [CrossRef]
11. T. Shintaku, A. Tate, and S. Mino, “Ce-substituted yttrium iron garnet films prepared on Gd3Sc2Ga3O12 garnet substrates by sputter epitaxy,” Appl. Phys. Lett. 71(12), 1640–1642 (1997). [CrossRef]
12. T. Shintaku, “Integrated optical isolator based on nonreciprocal higher-order mode conversion,” Appl. Phys. Lett. 66(21), 2789–2791 (1995). [CrossRef]
13. M. Gomi, H. Furuyama, and M. Abe, “Strong magneto-optical enhancement in highly Ce-substituted iron garnet films prepared by sputtering,” J. Appl. Phys. 70(11), 7065 (1991). [CrossRef]
14. M. Gomi, K. Satoh, H. Furuyama, and M. Abe, “Sputter deposition of Ce-substituted iron garnet films with giant magneto-optical effect,” J. Magn. Soc. Jpn. 13(2), 163–166 (1989). [CrossRef]
15. M. Gomi, K. Satoh, and M. Abe, “Giant Faraday rotation of Ce-substituted YIG films epitaxially grown by RF sputtering,” Jpn. J. Appl. Phys. 27(Part 2, No. 8), L1536–L1538 (1988). [CrossRef]
17. L. Bi, J. Hu, G. F. Dionne, L. Kimerling, and C. A. Ross, “Monolithic integration of chalcogenide glass/iron garnet waveguides and resonators for on-chip nonreciprocal photonic devices,” Proc. SPIE 7941, 5–10 (2011).
18. P. Grosseau, A. Bachiorrini, and B. Guilhot, “Preparation of polycrystalline yttrium iron garnet ceramics,” Powder Technol. 93(3), 247–251 (1997). [CrossRef]
19. N. B. Ibrahim, C. Edwards, and S. B. Palmer, “Pulsed laser ablation deposition of yttrium iron garnet and cerium-substituted YIG films,” J. Magn. Magn. Mater. 220(2-3), 183–194 (2000). [CrossRef]
20. T. Suzuki, “Magnetic and magneto-optic properties of rapid thermally crystallized garnet films,” J. Appl. Phys. 69(8), 4756–4760 (1991). [CrossRef]
21. K. Sato, Hikari-to-jiki (Asakura-shoten, Tokyo, 2007).
22. S. Kahl and A. M. Grishin, “Pulsed laser deposition of Y3Fe5O12 and Bi3Fe5O12 films on garnet substrates,” J. Appl. Phys. 93, 6945–6947 (2003). [CrossRef]
23. D. E. Aspnes and J. B. Theeten, “Optical properties of the interface between Si and its thermally grown oxide,” Phys. Rev. Lett. 43(14), 1046–1050 (1979). [CrossRef]
24. T. Nakano, H. Yuri, and U. Kihara, “Magneto-optical properties of YIG single crystal by TSFZ method,” IEEE Trans. Magn. 20(5), 986–988 (1984). [CrossRef]
25. W. Wettling, B. Andlauer, P. Koidl, J. Schneider, and W. Tolksdorf, “Optical absorption and Faraday rotation in yttrium iron garnet,” Phys. Status Solidi B 59(1), 63–70 (1973). [CrossRef]
26. W. A. Crossley, R. W. Cooper, J. L. Page, and R. P. van Stapele, “Faraday rotation in rare-earth iron garnets,” Phys. Rev. 181(2), 896–904 (1969). [CrossRef]
27. Y. Xu, J. H. Yang, and X. J. Zhang, “Quantum theory of the strong magneto-optical effect of Ce-substituted yttrium iron garnet,” Phys. Rev. B Condens. Matter 50(18), 13428–13434 (1994). [CrossRef] [PubMed]